Research On Wi-Fi Indoor Positioning Algorithm By Fusing Deep Feature And GBDT Feature

With the development of science and technology,mobile devices and wireless communication networks are widely used,and the demand for location-based services is increasing.Indoor and outdoor positioning technology is the cornerstone of location-based services.Outdoor positioning technology can locate targets through global satellite positioning systems in an outdoor environment and achieve high positioning accuracy.It is widely used in car navigation,medical services and rescue fields,and has great economic benefits and good market prospects.However,for the indoor environment,barriers such as walls,doors and windows,and interior furnishings can severely attenuate satellite signals,making it difficult for positioning systems to sense satellite signals.So outdoor positioning technology cannot be applied indoors.At this stage,Wi-Fi(Wireless Fidelity)networks are widely deployed in various places and most mobile devices such as mobile phones,tablets and laptops have embedded Wi-Fi chips.Wi-Fi positioning technology has become the most promising indoor positioning technology in the field of indoor positioning,which has aroused widespread concern in academia and industry.However,it is affected by many factors such as co-channel interference,complex and varied indoor environment and moving crowd.The RSS(Received Signal Strength)signal has severe volatility,which seriously affects the indoor positioning accuracy,and brings many problems to the Wi-Fi indoor positioning technology based on the fingerprint positioning algorithm.In this paper,the Wi-Fi indoor positioning technology based on fingerprint location algorithm is deeply studied,and the main problems affecting the positioning performance are pointed out.In view of the shortcomings of fingerprint localization algorithm,the theory of machine learning related to multi-task learning,feature extraction and feature fusion is fully utilized.A feature extraction algorithm based on multi-task learning and a precise location algorithm based on feature fusion are proposed,which improves the positioning technology based on fingerprint location algorithm to some extent.The main research work of this paper is summarized as follows:(1)Aiming at the volatility and high-dimensional sparsity of Wi-Fi signals,this paper proposes a feature extraction algorithm JMT-SDAE(Joint Multi-Task Stacked Denoising Auto-Encoder)based on multi-task learning to perform data dimensionality reduction and feature extraction on fingerprint data.The algorithm is verified using the store location dataset provided by the Tianchi Big Data Competition platform and compared with other algorithms.The experimental results show that the proposed JMT-SDAE algorithm has better ability to extract features and improve positioning accuracy.(2)In order to further improve the accuracy of Wi-Fi indoor positioning,this paper also proposes a hybrid model,which combines the deep features extracted by JMT-SDAE with GBDT(Gradient Boosting Decision Tree)features to construct a hybrid model to achieve accurate indoor positioning.The algorithm is verified using the store location dataset provided by the Tianchi Big Data Competition platform and compared with other algorithms.The experimental results show that the hybrid model proposed in this paper significantly improves the positioning accuracy.